Negative risk free rate in second order computation

Hi, all,
I encountered a very wired problem when I compute my model using Dynare and compute moments using simulation data.
The problem is that:
if I use order 1, in the simulation data, I get the mean of riskfree rate rougly equal to my targeted value, 2% per year.
However, if I change the order to 2, in the simulation data, the mean of riskfree rate becomes nagetiave, with different parameter values, it can be from -2% to -5%.
I have tried to adjust the risk aversion parameter, IES parameter, the persistence of TFP shock, but the results in order 2 are still negative.

Do any one know what’s going on? Is there anyway to make it positive?
Thank you in advance.

I would need to see the file.